Re: US Enqueue , latch row object cache - performance issue

From: orapc <puravc_at_gmail.com>
Date: Thu, 20 Aug 2009 11:44:37 -0700 (PDT)
Message-ID: <a84b5cdc-c974-4b4c-8ea4-5a997380b9fe_at_v15g2000prn.googlegroups.com>



On Aug 11, 11:35 am, Grzegorz <grzegor..._at_interia.pl> wrote:
> Charles Hooper pisze:
>
> > Greg,
>
> > Take a look at Metalink Doc ID 463136.1 "POOR PERFORMANCE DUE TO HUGE
> > WAITS ONUSENQUEUE" which applies to versions 9.0.1.4 to 9.2.0.8.
> > Doc ID 301432.1 "Performance Slowdown During Heavy Undo Segment
> > Onlining" states that it applies to 9.2.0.5, but the work-around is
> > almost the same as for Doc ID 463136.1.
>
> Thanks, that looks promising :).
> Regards.
> Greg

Greg,
Did you set the event as mentioned in the metalink doc and if yes then did it help? In our case it is 9.2.0.6 RAC and we were facing this issue. So we set the event with level as 2 on both instances and bounced both instances. However 2 days later we again faced the same issue. So it either means that we were not facing that problem in the first place OR it means that setting the event did not help OR something else. However, every peice of information for the US enqueue that is available either on metalink or internet states that it happens only when onling/offlining by SMON. If it is so then the moment I see US enqueues in statspack then it means we are/were certainly facing the very issue and if so then setting the event should have helped. So I am not able to understand what next to be done. Whenever we see US enqueue, we also see huge waits on HW enqueue.

The only thing is that metalink doc id 463136.1 suggests to set the event at level 1 whereas doc id 301432.1 suggests to set the event at level 2. If that means that if the version is 9.2.0.5 then level 2 is correct and for the rest it has to be level 1 then it could be a problem because we have set it at level 2 whereas our version is 9.2.0.6.

Thanks. Received on Thu Aug 20 2009 - 13:44:37 CDT

Original text of this message